A novel multi-objective optimization algorithm for Pareto design of a fuzzy full state feedback linearization controller applied on a ball and wheel system

The control and stabilization of a ball and wheel system around the equilibrium point are challenging tasks because it is an underactuated, nonlinear, and open-loop unstable plant. In this paper, Pareto design of a Fuzzy Full State Feedback Linearization Controller (FFSFLC) for the ball and wheel system based upon a novel multi-objective optimization algorithm is introduced. To this end, at first, a full state feedback linearization approach is employed to stabilize the dynamics of the system. Next, appropriate fuzzy systems are determined to tune the control gains. Then, a new multi-objective optimization algorithm is utilized to promote the proposed control scheme. This optimization algorithm is a combination of Simulated Annealing (SA) and Artificial Bee Colony (ABC) approaches benefiting advantages of the non-dominated Pareto solutions. To evaluate the capabilities of the suggested algorithm, its optimal solutions of several standard test functions are compared with those of five renowned multi-objective optimization algorithms. The results confirm that the proposed hybrid algorithm yields closer non-dominated Pareto solutions to the true optimal Pareto front with shorter runtimes than other algorithms. After selecting proper objective functions, multi-objective optimization of FFSFLC for the ball and wheel system is performed, and the results are compared with previous works. Simulations illustrate that the proposed strategies can accurately converge the system states to the desired conditions and yield superior robustness against disturbance signals in comparison with former studies.

Evolutionary Multi-Objective One-Shot Filter Pruning for Designing Lightweight Convolutional Neural Network

Deep neural networks have achieved significant development and wide applications for their amazing performance. However, their complex structure, high computation and storage resource limit their applications in mobile or embedding devices such as sensor platforms. Neural network pruning is an efficient way to design a lightweight model from a well-trained complex deep neural network. In this paper, we propose an evolutionary multi-objective one-shot filter pruning method for designing a lightweight convolutional neural network. Firstly, unlike some famous iterative pruning methods, a one-shot pruning framework only needs to perform filter pruning and model fine-tuning once. Moreover, we built a constraint multi-objective filter pruning problem in which two objectives represent the filter pruning ratio and the accuracy of the pruned convolutional neural network, respectively. A non-dominated sorting-based evolutionary multi-objective algorithm was used to solve the filter pruning problem, and it provides a set of Pareto solutions which consists of a series of different trade-off pruned models. Finally, some models are uniformly selected from the set of Pareto solutions to be fine-tuned as the output of our method. The effectiveness of our method was demonstrated in experimental studies on four designed models, LeNet and AlexNet. Our method can prune over 85%, 82%, 75%, 65%, 91% and 68% filters with little accuracy loss on four designed models, LeNet and AlexNet, respectively.

A Hybrid Multi-Objective Optimization Method Based on NSGA-II Algorithm and Entropy Weighted TOPSIS for Lightweight Design of Dump Truck Carriage

The lightweight design of vehicle components is regarded as a complex optimization problem, which usually needs to achieve two or more optimization objectives. It can be firstly solved by a multi-objective optimization algorithm for generating Pareto solutions, before then seeking the optimal design. However, it is difficult to determine the optimal design for lack of engineering knowledge about ideal and nadir values. Therefore, this paper proposes a multi-objective optimization procedure combined with the NSGA-II algorithm with entropy weighted TOPSIS for the lightweight design of the dump truck carriage. The finite element model of the dump truck carriage was firstly developed for modal analysis under unconstrained free state and strength analysis under the full load and lifting conditions. On this basis, the multi-objective lightweight optimization of the dump truck carriage was carried out based on the Kriging surrogate model and the NSGA-II algorithm. Then, the entropy weight TOPSIS method was employed to select the optimal design of the dump truck from Pareto solutions. The results show that the optimized dump truck carriage achieves a remarkable mass reduction of 81 kg, as much as 3.7%, while its first-order natural frequency and strength performance are slightly improved compared with the original model. Accordingly, the proposed procedure provides an effective way for vehicle lightweight design.

Dynamic immune cooperative scheduling of agricultural machineries

Considering the low flexibility and efficiency of the scheduling problem, an improved multi-objective immune algorithm with non-dominated neighbor-based selection and Tabu search (NNITSA) is proposed. A novel Tabu search algorithm (TSA)-based operator is introduced in both the local search and mutation stage, which improves the climbing performance of the NNTSA. Special local search strategies can prevent the algorithm from being caught in the optimal solution. In addition, considering the time costs of the TSA, an adapted mutation strategy is proposed to operate the TSA mutation according to the scale of Pareto solutions. Random mutations may be applied to other conditions. Then, a robust evaluation is adopted to choose an appropriate solution from the obtained Pareto solutions set. NNITSA is used to solve the problems of static partitioning optimization and dynamic cross-regional co-operative scheduling of agricultural machinery. The simulation results show that NNITSA outperforms the other two algorithms, NNIA and NSGA-II. The performance indicator C-metric also shows significant improvements in the efficiency of optimizing search.

Solving a multicriteria scheduling problem using a genetic algorithm

The work is devoted to the multiobjective task of scheduling, in which a given set of works must be performed by several performers of different productivity. A certain number of bonuses is accrued for the work performed by the respective executor, which depends on the time of work performance. The criteria for evaluating the schedule are the total time of all jobs and the amount of bonuses spent. In the research the main approaches to solving multiobjective optimization problems were analyzed, based on which the Pareto approach was chosen. The genetic algorithm was chosen as the algorithm. The purpose of this work is to increase the efficiency of solving multicriteria optimization problems by implementing a heuristic algorithm and increase its speed. The tasks of the work are to determine the advantages and disadvantages of the approaches used to solve multicriteria optimization problems, to develop a genetic algorithm for solving the multicriteria scheduling problem and to study its efficiency. Operators of the genetic algorithm have been developed, which take into account the peculiarities of the researched problem and allow to obtain Pareto solutions in the process of work. Due to the introduction of parallel calculations in the implementation of the genetic algorithm, it was possible to increase its speed compared to the conventional version. The developed algorithm can be used in solving the problem of optimal allocation of resources, which is part of the system of accrual of bonuses to employees.

An Efficient Multi-Objective Robust Optimization Method by Sequentially Searching From Nominal Pareto Solutions

Multi-objective optimization problems (MOOPs) with uncertainties are common in engineering design. To find robust Pareto fronts, multi-objective robust optimization (MORO) methods with inner–outer optimization structures usually have high computational complexity, which is a critical issue. Generally, in design problems, robust Pareto solutions lie somewhere closer to nominal Pareto points compared with randomly initialized points. The searching process for robust solutions could be more efficient if starting from nominal Pareto points. We propose a new method sequentially approaching to the robust Pareto front (SARPF) from the nominal Pareto points where MOOPs with uncertainties are solved in two stages. The deterministic optimization problem and robustness metric optimization are solved in the first stage, where nominal Pareto solutions and the robust-most solutions are identified, respectively. In the second stage, a new single-objective robust optimization problem is formulated to find the robust Pareto solutions starting from the nominal Pareto points in the region between the nominal Pareto front and robust-most points. The proposed SARPF method can reduce a significant amount of computational time since the optimization process can be performed in parallel at each stage. Vertex estimation is also applied to approximate the worst-case uncertain parameter values, which can reduce computational efforts further. The global solvers, NSGA-II for multi-objective cases and genetic algorithm (GA) for single-objective cases, are used in corresponding optimization processes. Three examples with the comparison with results from the previous method are presented to demonstrate the applicability and efficiency of the proposed method.